2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
20 #include <linux/pagemap.h>
21 #include <linux/highmem.h>
22 #include <linux/time.h>
23 #include <linux/init.h>
24 #include <linux/string.h>
25 #include <linux/smp_lock.h>
26 #include <linux/backing-dev.h>
27 #include <linux/mpage.h>
28 #include <linux/swap.h>
29 #include <linux/writeback.h>
30 #include <linux/statfs.h>
31 #include <linux/compat.h>
32 #include <linux/version.h>
35 #include "transaction.h"
36 #include "btrfs_inode.h"
38 #include "print-tree.h"
42 static int btrfs_copy_from_user(loff_t pos
, int num_pages
, int write_bytes
,
43 struct page
**prepared_pages
,
44 const char __user
* buf
)
48 int offset
= pos
& (PAGE_CACHE_SIZE
- 1);
50 for (i
= 0; i
< num_pages
&& write_bytes
> 0; i
++, offset
= 0) {
51 size_t count
= min_t(size_t,
52 PAGE_CACHE_SIZE
- offset
, write_bytes
);
53 struct page
*page
= prepared_pages
[i
];
54 fault_in_pages_readable(buf
, count
);
56 /* Copy data from userspace to the current page */
58 page_fault
= __copy_from_user(page_address(page
) + offset
,
60 /* Flush processor's dcache for this page */
61 flush_dcache_page(page
);
69 return page_fault
? -EFAULT
: 0;
72 static void btrfs_drop_pages(struct page
**pages
, size_t num_pages
)
75 for (i
= 0; i
< num_pages
; i
++) {
78 ClearPageChecked(pages
[i
]);
79 unlock_page(pages
[i
]);
80 mark_page_accessed(pages
[i
]);
81 page_cache_release(pages
[i
]);
85 static int noinline
insert_inline_extent(struct btrfs_trans_handle
*trans
,
86 struct btrfs_root
*root
, struct inode
*inode
,
87 u64 offset
, size_t size
,
88 struct page
**pages
, size_t page_offset
,
92 struct btrfs_path
*path
;
93 struct extent_buffer
*leaf
;
96 struct btrfs_file_extent_item
*ei
;
104 path
= btrfs_alloc_path();
108 btrfs_set_trans_block_group(trans
, inode
);
110 key
.objectid
= inode
->i_ino
;
112 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
114 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 1);
120 struct btrfs_key found_key
;
122 if (path
->slots
[0] == 0)
126 leaf
= path
->nodes
[0];
127 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
129 if (found_key
.objectid
!= inode
->i_ino
)
132 if (found_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
134 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
135 struct btrfs_file_extent_item
);
137 if (btrfs_file_extent_type(leaf
, ei
) !=
138 BTRFS_FILE_EXTENT_INLINE
) {
141 btrfs_item_key_to_cpu(leaf
, &key
, path
->slots
[0]);
148 leaf
= path
->nodes
[0];
149 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
150 struct btrfs_file_extent_item
);
152 if (btrfs_file_extent_type(leaf
, ei
) !=
153 BTRFS_FILE_EXTENT_INLINE
) {
155 btrfs_print_leaf(root
, leaf
);
156 printk("found wasn't inline offset %Lu inode %lu\n",
157 offset
, inode
->i_ino
);
160 found_size
= btrfs_file_extent_inline_len(leaf
,
161 btrfs_item_nr(leaf
, path
->slots
[0]));
162 found_end
= key
.offset
+ found_size
;
164 if (found_end
< offset
+ size
) {
165 btrfs_release_path(root
, path
);
166 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
167 offset
+ size
- found_end
, 1);
170 ret
= btrfs_extend_item(trans
, root
, path
,
171 offset
+ size
- found_end
);
176 leaf
= path
->nodes
[0];
177 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
178 struct btrfs_file_extent_item
);
179 inode
->i_blocks
+= (offset
+ size
- found_end
) >> 9;
181 if (found_end
< offset
) {
182 ptr
= btrfs_file_extent_inline_start(ei
) + found_size
;
183 memset_extent_buffer(leaf
, 0, ptr
, offset
- found_end
);
187 btrfs_release_path(root
, path
);
188 datasize
= offset
+ size
- key
.offset
;
189 inode
->i_blocks
+= datasize
>> 9;
190 datasize
= btrfs_file_extent_calc_inline_size(datasize
);
191 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
195 printk("got bad ret %d\n", ret
);
198 leaf
= path
->nodes
[0];
199 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
200 struct btrfs_file_extent_item
);
201 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
202 btrfs_set_file_extent_type(leaf
, ei
, BTRFS_FILE_EXTENT_INLINE
);
204 ptr
= btrfs_file_extent_inline_start(ei
) + offset
- key
.offset
;
210 kaddr
= kmap_atomic(page
, KM_USER0
);
211 cur_size
= min_t(size_t, PAGE_CACHE_SIZE
- page_offset
, size
);
212 write_extent_buffer(leaf
, kaddr
+ page_offset
, ptr
, cur_size
);
213 kunmap_atomic(kaddr
, KM_USER0
);
217 if (i
>= num_pages
) {
218 printk("i %d num_pages %d\n", i
, num_pages
);
222 btrfs_mark_buffer_dirty(leaf
);
224 btrfs_free_path(path
);
228 static int noinline
dirty_and_release_pages(struct btrfs_trans_handle
*trans
,
229 struct btrfs_root
*root
,
238 struct inode
*inode
= fdentry(file
)->d_inode
;
239 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
243 u64 end_of_last_block
;
244 u64 end_pos
= pos
+ write_bytes
;
247 loff_t isize
= i_size_read(inode
);
249 start_pos
= pos
& ~((u64
)root
->sectorsize
- 1);
250 num_bytes
= (write_bytes
+ pos
- start_pos
+
251 root
->sectorsize
- 1) & ~((u64
)root
->sectorsize
- 1);
253 end_of_last_block
= start_pos
+ num_bytes
- 1;
255 lock_extent(io_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
256 trans
= btrfs_join_transaction(root
, 1);
261 btrfs_set_trans_block_group(trans
, inode
);
264 if ((end_of_last_block
& 4095) == 0) {
265 printk("strange end of last %Lu %zu %Lu\n", start_pos
, write_bytes
, end_of_last_block
);
267 set_extent_uptodate(io_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
269 /* FIXME...EIEIO, ENOSPC and more */
270 /* insert any holes we need to create */
271 if (isize
< start_pos
) {
272 u64 last_pos_in_file
;
274 u64 mask
= root
->sectorsize
- 1;
275 last_pos_in_file
= (isize
+ mask
) & ~mask
;
276 hole_size
= (start_pos
- last_pos_in_file
+ mask
) & ~mask
;
278 btrfs_wait_ordered_range(inode
, last_pos_in_file
,
279 last_pos_in_file
+ hole_size
);
280 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
281 err
= btrfs_drop_extents(trans
, root
, inode
,
283 last_pos_in_file
+ hole_size
,
289 err
= btrfs_insert_file_extent(trans
, root
,
293 btrfs_drop_extent_cache(inode
, last_pos_in_file
,
294 last_pos_in_file
+ hole_size
-1);
295 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
296 btrfs_check_file(root
, inode
);
303 * either allocate an extent for the new bytes or setup the key
304 * to show we are doing inline data in the extent
306 inline_size
= end_pos
;
307 if (isize
>= BTRFS_MAX_INLINE_DATA_SIZE(root
) ||
308 inline_size
> root
->fs_info
->max_inline
||
309 (inline_size
& (root
->sectorsize
-1)) == 0 ||
310 inline_size
>= BTRFS_MAX_INLINE_DATA_SIZE(root
)) {
311 /* check for reserved extents on each page, we don't want
312 * to reset the delalloc bit on things that already have
315 btrfs_set_extent_delalloc(inode
, start_pos
, end_of_last_block
);
316 for (i
= 0; i
< num_pages
; i
++) {
317 struct page
*p
= pages
[i
];
324 /* step one, delete the existing extents in this range */
325 aligned_end
= (pos
+ write_bytes
+ root
->sectorsize
- 1) &
326 ~((u64
)root
->sectorsize
- 1);
327 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
328 err
= btrfs_drop_extents(trans
, root
, inode
, start_pos
,
329 aligned_end
, aligned_end
, &hint_byte
);
332 if (isize
> inline_size
)
333 inline_size
= min_t(u64
, isize
, aligned_end
);
334 inline_size
-= start_pos
;
335 err
= insert_inline_extent(trans
, root
, inode
, start_pos
,
336 inline_size
, pages
, 0, num_pages
);
337 btrfs_drop_extent_cache(inode
, start_pos
, aligned_end
- 1);
339 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
342 * an ugly way to do all the prop accounting around
343 * the page bits and mapping tags
345 set_page_writeback(pages
[0]);
346 end_page_writeback(pages
[0]);
349 if (end_pos
> isize
) {
350 i_size_write(inode
, end_pos
);
352 BTRFS_I(inode
)->disk_i_size
= end_pos
;
353 btrfs_update_inode(trans
, root
, inode
);
356 err
= btrfs_end_transaction(trans
, root
);
358 unlock_extent(io_tree
, start_pos
, end_of_last_block
, GFP_NOFS
);
362 int btrfs_drop_extent_cache(struct inode
*inode
, u64 start
, u64 end
)
364 struct extent_map
*em
;
365 struct extent_map
*split
= NULL
;
366 struct extent_map
*split2
= NULL
;
367 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
368 u64 len
= end
- start
+ 1;
372 WARN_ON(end
< start
);
373 if (end
== (u64
)-1) {
379 split
= alloc_extent_map(GFP_NOFS
);
381 split2
= alloc_extent_map(GFP_NOFS
);
383 spin_lock(&em_tree
->lock
);
384 em
= lookup_extent_mapping(em_tree
, start
, len
);
386 spin_unlock(&em_tree
->lock
);
389 clear_bit(EXTENT_FLAG_PINNED
, &em
->flags
);
390 remove_extent_mapping(em_tree
, em
);
392 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
&&
394 split
->start
= em
->start
;
395 split
->len
= start
- em
->start
;
396 split
->block_start
= em
->block_start
;
397 split
->bdev
= em
->bdev
;
398 split
->flags
= em
->flags
;
399 ret
= add_extent_mapping(em_tree
, split
);
401 free_extent_map(split
);
405 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
&&
406 testend
&& em
->start
+ em
->len
> start
+ len
) {
407 u64 diff
= start
+ len
- em
->start
;
409 split
->start
= start
+ len
;
410 split
->len
= em
->start
+ em
->len
- (start
+ len
);
411 split
->bdev
= em
->bdev
;
412 split
->flags
= em
->flags
;
414 split
->block_start
= em
->block_start
+ diff
;
416 ret
= add_extent_mapping(em_tree
, split
);
418 free_extent_map(split
);
421 spin_unlock(&em_tree
->lock
);
425 /* once for the tree*/
429 free_extent_map(split
);
431 free_extent_map(split2
);
435 int btrfs_check_file(struct btrfs_root
*root
, struct inode
*inode
)
439 struct btrfs_path
*path
;
440 struct btrfs_key found_key
;
441 struct extent_buffer
*leaf
;
442 struct btrfs_file_extent_item
*extent
;
451 path
= btrfs_alloc_path();
452 ret
= btrfs_lookup_file_extent(NULL
, root
, path
, inode
->i_ino
,
455 nritems
= btrfs_header_nritems(path
->nodes
[0]);
456 if (path
->slots
[0] >= nritems
) {
457 ret
= btrfs_next_leaf(root
, path
);
460 nritems
= btrfs_header_nritems(path
->nodes
[0]);
462 slot
= path
->slots
[0];
463 leaf
= path
->nodes
[0];
464 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
465 if (found_key
.objectid
!= inode
->i_ino
)
467 if (found_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
470 if (found_key
.offset
< last_offset
) {
472 btrfs_print_leaf(root
, leaf
);
473 printk("inode %lu found offset %Lu expected %Lu\n",
474 inode
->i_ino
, found_key
.offset
, last_offset
);
478 extent
= btrfs_item_ptr(leaf
, slot
,
479 struct btrfs_file_extent_item
);
480 found_type
= btrfs_file_extent_type(leaf
, extent
);
481 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
482 extent_end
= found_key
.offset
+
483 btrfs_file_extent_num_bytes(leaf
, extent
);
484 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
485 struct btrfs_item
*item
;
486 item
= btrfs_item_nr(leaf
, slot
);
487 extent_end
= found_key
.offset
+
488 btrfs_file_extent_inline_len(leaf
, item
);
489 extent_end
= (extent_end
+ root
->sectorsize
- 1) &
490 ~((u64
)root
->sectorsize
-1 );
492 last_offset
= extent_end
;
495 if (0 && last_offset
< inode
->i_size
) {
497 btrfs_print_leaf(root
, leaf
);
498 printk("inode %lu found offset %Lu size %Lu\n", inode
->i_ino
,
499 last_offset
, inode
->i_size
);
504 btrfs_free_path(path
);
510 * this is very complex, but the basic idea is to drop all extents
511 * in the range start - end. hint_block is filled in with a block number
512 * that would be a good hint to the block allocator for this file.
514 * If an extent intersects the range but is not entirely inside the range
515 * it is either truncated or split. Anything entirely inside the range
516 * is deleted from the tree.
518 int btrfs_drop_extents(struct btrfs_trans_handle
*trans
,
519 struct btrfs_root
*root
, struct inode
*inode
,
520 u64 start
, u64 end
, u64 inline_limit
, u64
*hint_byte
)
523 u64 search_start
= start
;
524 struct extent_buffer
*leaf
;
525 struct btrfs_file_extent_item
*extent
;
526 struct btrfs_path
*path
;
527 struct btrfs_key key
;
528 struct btrfs_file_extent_item old
;
538 btrfs_drop_extent_cache(inode
, start
, end
- 1);
540 path
= btrfs_alloc_path();
545 btrfs_release_path(root
, path
);
546 ret
= btrfs_lookup_file_extent(trans
, root
, path
, inode
->i_ino
,
551 if (path
->slots
[0] == 0) {
563 leaf
= path
->nodes
[0];
564 slot
= path
->slots
[0];
566 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
567 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
&&
571 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
572 key
.objectid
!= inode
->i_ino
) {
576 search_start
= key
.offset
;
579 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
580 extent
= btrfs_item_ptr(leaf
, slot
,
581 struct btrfs_file_extent_item
);
582 found_type
= btrfs_file_extent_type(leaf
, extent
);
583 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
585 btrfs_file_extent_disk_bytenr(leaf
,
588 *hint_byte
= extent_end
;
590 extent_end
= key
.offset
+
591 btrfs_file_extent_num_bytes(leaf
, extent
);
593 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
594 struct btrfs_item
*item
;
595 item
= btrfs_item_nr(leaf
, slot
);
597 extent_end
= key
.offset
+
598 btrfs_file_extent_inline_len(leaf
, item
);
601 extent_end
= search_start
;
604 /* we found nothing we can drop */
605 if ((!found_extent
&& !found_inline
) ||
606 search_start
>= extent_end
) {
609 nritems
= btrfs_header_nritems(leaf
);
610 if (slot
>= nritems
- 1) {
611 nextret
= btrfs_next_leaf(root
, path
);
622 u64 mask
= root
->sectorsize
- 1;
623 search_start
= (extent_end
+ mask
) & ~mask
;
625 search_start
= extent_end
;
626 if (end
<= extent_end
&& start
>= key
.offset
&& found_inline
) {
627 *hint_byte
= EXTENT_MAP_INLINE
;
630 if (end
< extent_end
&& end
>= key
.offset
) {
633 btrfs_file_extent_disk_bytenr(leaf
, extent
);
635 btrfs_file_extent_disk_num_bytes(leaf
,
637 read_extent_buffer(leaf
, &old
,
638 (unsigned long)extent
,
640 if (disk_bytenr
!= 0) {
641 ret
= btrfs_inc_extent_ref(trans
, root
,
642 disk_bytenr
, disk_num_bytes
,
643 root
->root_key
.objectid
,
650 if (found_inline
&& start
<= key
.offset
)
653 /* truncate existing extent */
654 if (start
> key
.offset
) {
658 WARN_ON(start
& (root
->sectorsize
- 1));
660 new_num
= start
- key
.offset
;
661 old_num
= btrfs_file_extent_num_bytes(leaf
,
664 btrfs_file_extent_disk_bytenr(leaf
,
666 if (btrfs_file_extent_disk_bytenr(leaf
,
668 dec_i_blocks(inode
, old_num
- new_num
);
670 btrfs_set_file_extent_num_bytes(leaf
, extent
,
672 btrfs_mark_buffer_dirty(leaf
);
673 } else if (key
.offset
< inline_limit
&&
674 (end
> extent_end
) &&
675 (inline_limit
< extent_end
)) {
677 new_size
= btrfs_file_extent_calc_inline_size(
678 inline_limit
- key
.offset
);
679 dec_i_blocks(inode
, (extent_end
- key
.offset
) -
680 (inline_limit
- key
.offset
));
681 btrfs_truncate_item(trans
, root
, path
,
685 /* delete the entire extent */
688 u64 disk_num_bytes
= 0;
689 u64 extent_num_bytes
= 0;
693 root_gen
= btrfs_header_generation(leaf
);
694 root_owner
= btrfs_header_owner(leaf
);
697 btrfs_file_extent_disk_bytenr(leaf
,
700 btrfs_file_extent_disk_num_bytes(leaf
,
703 btrfs_file_extent_num_bytes(leaf
, extent
);
705 btrfs_file_extent_disk_bytenr(leaf
,
708 ret
= btrfs_del_item(trans
, root
, path
);
709 /* TODO update progress marker and return */
711 btrfs_release_path(root
, path
);
713 if (found_extent
&& disk_bytenr
!= 0) {
714 dec_i_blocks(inode
, extent_num_bytes
);
715 ret
= btrfs_free_extent(trans
, root
,
719 root_gen
, inode
->i_ino
,
724 if (!bookend
&& search_start
>= end
) {
731 if (bookend
&& found_inline
&& start
<= key
.offset
) {
733 new_size
= btrfs_file_extent_calc_inline_size(
735 dec_i_blocks(inode
, (extent_end
- key
.offset
) -
737 btrfs_truncate_item(trans
, root
, path
, new_size
, 0);
739 /* create bookend, splitting the extent in two */
740 if (bookend
&& found_extent
) {
741 struct btrfs_key ins
;
742 ins
.objectid
= inode
->i_ino
;
744 btrfs_set_key_type(&ins
, BTRFS_EXTENT_DATA_KEY
);
745 btrfs_release_path(root
, path
);
746 ret
= btrfs_insert_empty_item(trans
, root
, path
, &ins
,
749 leaf
= path
->nodes
[0];
751 btrfs_print_leaf(root
, leaf
);
752 printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret
, ins
.objectid
, ins
.type
, ins
.offset
, start
, end
, key
.offset
, extent_end
, keep
);
755 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
756 struct btrfs_file_extent_item
);
757 write_extent_buffer(leaf
, &old
,
758 (unsigned long)extent
, sizeof(old
));
760 btrfs_set_file_extent_offset(leaf
, extent
,
761 le64_to_cpu(old
.offset
) + end
- key
.offset
);
762 WARN_ON(le64_to_cpu(old
.num_bytes
) <
764 btrfs_set_file_extent_num_bytes(leaf
, extent
,
766 btrfs_set_file_extent_type(leaf
, extent
,
767 BTRFS_FILE_EXTENT_REG
);
769 btrfs_mark_buffer_dirty(path
->nodes
[0]);
770 if (le64_to_cpu(old
.disk_bytenr
) != 0) {
772 btrfs_file_extent_num_bytes(leaf
,
780 btrfs_free_path(path
);
781 btrfs_check_file(root
, inode
);
786 * this gets pages into the page cache and locks them down
788 static int prepare_pages(struct btrfs_root
*root
, struct file
*file
,
789 struct page
**pages
, size_t num_pages
,
790 loff_t pos
, unsigned long first_index
,
791 unsigned long last_index
, size_t write_bytes
)
794 unsigned long index
= pos
>> PAGE_CACHE_SHIFT
;
795 struct inode
*inode
= fdentry(file
)->d_inode
;
800 start_pos
= pos
& ~((u64
)root
->sectorsize
- 1);
801 last_pos
= ((u64
)index
+ num_pages
) << PAGE_CACHE_SHIFT
;
803 memset(pages
, 0, num_pages
* sizeof(struct page
*));
805 for (i
= 0; i
< num_pages
; i
++) {
806 pages
[i
] = grab_cache_page(inode
->i_mapping
, index
+ i
);
811 wait_on_page_writeback(pages
[i
]);
813 if (start_pos
< inode
->i_size
) {
814 struct btrfs_ordered_extent
*ordered
;
815 lock_extent(&BTRFS_I(inode
)->io_tree
,
816 start_pos
, last_pos
- 1, GFP_NOFS
);
817 ordered
= btrfs_lookup_first_ordered_extent(inode
, last_pos
-1);
819 ordered
->file_offset
+ ordered
->len
> start_pos
&&
820 ordered
->file_offset
< last_pos
) {
821 btrfs_put_ordered_extent(ordered
);
822 unlock_extent(&BTRFS_I(inode
)->io_tree
,
823 start_pos
, last_pos
- 1, GFP_NOFS
);
824 for (i
= 0; i
< num_pages
; i
++) {
825 unlock_page(pages
[i
]);
826 page_cache_release(pages
[i
]);
828 btrfs_wait_ordered_range(inode
, start_pos
,
829 last_pos
- start_pos
);
833 btrfs_put_ordered_extent(ordered
);
835 clear_extent_bits(&BTRFS_I(inode
)->io_tree
, start_pos
,
836 last_pos
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
,
838 unlock_extent(&BTRFS_I(inode
)->io_tree
,
839 start_pos
, last_pos
- 1, GFP_NOFS
);
841 for (i
= 0; i
< num_pages
; i
++) {
842 clear_page_dirty_for_io(pages
[i
]);
843 set_page_extent_mapped(pages
[i
]);
844 WARN_ON(!PageLocked(pages
[i
]));
849 static ssize_t
btrfs_file_write(struct file
*file
, const char __user
*buf
,
850 size_t count
, loff_t
*ppos
)
854 ssize_t num_written
= 0;
857 struct inode
*inode
= fdentry(file
)->d_inode
;
858 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
859 struct page
**pages
= NULL
;
861 struct page
*pinned
[2];
862 unsigned long first_index
;
863 unsigned long last_index
;
865 nrptrs
= min((count
+ PAGE_CACHE_SIZE
- 1) / PAGE_CACHE_SIZE
,
866 PAGE_CACHE_SIZE
/ (sizeof(struct page
*)));
873 vfs_check_frozen(inode
->i_sb
, SB_FREEZE_WRITE
);
874 current
->backing_dev_info
= inode
->i_mapping
->backing_dev_info
;
875 err
= generic_write_checks(file
, &pos
, &count
, S_ISBLK(inode
->i_mode
));
880 #ifdef REMOVE_SUID_PATH
881 err
= remove_suid(&file
->f_path
);
883 # if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
884 err
= file_remove_suid(file
);
886 err
= remove_suid(fdentry(file
));
891 file_update_time(file
);
893 pages
= kmalloc(nrptrs
* sizeof(struct page
*), GFP_KERNEL
);
895 mutex_lock(&inode
->i_mutex
);
896 first_index
= pos
>> PAGE_CACHE_SHIFT
;
897 last_index
= (pos
+ count
) >> PAGE_CACHE_SHIFT
;
900 * if this is a nodatasum mount, force summing off for the inode
901 * all the time. That way a later mount with summing on won't
904 if (btrfs_test_opt(root
, NODATASUM
))
905 btrfs_set_flag(inode
, NODATASUM
);
908 * there are lots of better ways to do this, but this code
909 * makes sure the first and last page in the file range are
910 * up to date and ready for cow
912 if ((pos
& (PAGE_CACHE_SIZE
- 1))) {
913 pinned
[0] = grab_cache_page(inode
->i_mapping
, first_index
);
914 if (!PageUptodate(pinned
[0])) {
915 ret
= btrfs_readpage(NULL
, pinned
[0]);
917 wait_on_page_locked(pinned
[0]);
919 unlock_page(pinned
[0]);
922 if ((pos
+ count
) & (PAGE_CACHE_SIZE
- 1)) {
923 pinned
[1] = grab_cache_page(inode
->i_mapping
, last_index
);
924 if (!PageUptodate(pinned
[1])) {
925 ret
= btrfs_readpage(NULL
, pinned
[1]);
927 wait_on_page_locked(pinned
[1]);
929 unlock_page(pinned
[1]);
934 size_t offset
= pos
& (PAGE_CACHE_SIZE
- 1);
935 size_t write_bytes
= min(count
, nrptrs
*
936 (size_t)PAGE_CACHE_SIZE
-
938 size_t num_pages
= (write_bytes
+ PAGE_CACHE_SIZE
- 1) >>
941 WARN_ON(num_pages
> nrptrs
);
942 memset(pages
, 0, sizeof(pages
));
944 ret
= btrfs_check_free_space(root
, write_bytes
, 0);
948 ret
= prepare_pages(root
, file
, pages
, num_pages
,
949 pos
, first_index
, last_index
,
954 ret
= btrfs_copy_from_user(pos
, num_pages
,
955 write_bytes
, pages
, buf
);
957 btrfs_drop_pages(pages
, num_pages
);
961 ret
= dirty_and_release_pages(NULL
, root
, file
, pages
,
962 num_pages
, pos
, write_bytes
);
963 btrfs_drop_pages(pages
, num_pages
);
968 count
-= write_bytes
;
970 num_written
+= write_bytes
;
972 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, num_pages
);
973 if (num_pages
< (root
->leafsize
>> PAGE_CACHE_SHIFT
) + 1)
974 btrfs_btree_balance_dirty(root
, 1);
975 btrfs_throttle(root
);
979 mutex_unlock(&inode
->i_mutex
);
984 page_cache_release(pinned
[0]);
986 page_cache_release(pinned
[1]);
989 if (num_written
> 0 && ((file
->f_flags
& O_SYNC
) || IS_SYNC(inode
))) {
990 err
= sync_page_range(inode
, inode
->i_mapping
,
991 start_pos
, num_written
);
994 } else if (num_written
> 0 && (file
->f_flags
& O_DIRECT
)) {
995 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
996 do_sync_file_range(file
, start_pos
,
997 start_pos
+ num_written
- 1,
998 SYNC_FILE_RANGE_WRITE
|
999 SYNC_FILE_RANGE_WAIT_AFTER
);
1001 do_sync_mapping_range(inode
->i_mapping
, start_pos
,
1002 start_pos
+ num_written
- 1,
1003 SYNC_FILE_RANGE_WRITE
|
1004 SYNC_FILE_RANGE_WAIT_AFTER
);
1006 invalidate_mapping_pages(inode
->i_mapping
,
1007 start_pos
>> PAGE_CACHE_SHIFT
,
1008 (start_pos
+ num_written
- 1) >> PAGE_CACHE_SHIFT
);
1010 current
->backing_dev_info
= NULL
;
1011 return num_written
? num_written
: err
;
1014 int btrfs_release_file(struct inode
* inode
, struct file
* filp
)
1016 if (filp
->private_data
)
1017 btrfs_ioctl_trans_end(filp
);
1021 static int btrfs_sync_file(struct file
*file
,
1022 struct dentry
*dentry
, int datasync
)
1024 struct inode
*inode
= dentry
->d_inode
;
1025 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1027 struct btrfs_trans_handle
*trans
;
1030 * check the transaction that last modified this inode
1031 * and see if its already been committed
1033 if (!BTRFS_I(inode
)->last_trans
)
1036 mutex_lock(&root
->fs_info
->trans_mutex
);
1037 if (BTRFS_I(inode
)->last_trans
<=
1038 root
->fs_info
->last_trans_committed
) {
1039 BTRFS_I(inode
)->last_trans
= 0;
1040 mutex_unlock(&root
->fs_info
->trans_mutex
);
1043 mutex_unlock(&root
->fs_info
->trans_mutex
);
1046 * ok we haven't committed the transaction yet, lets do a commit
1048 if (file
->private_data
)
1049 btrfs_ioctl_trans_end(file
);
1051 trans
= btrfs_start_transaction(root
, 1);
1056 ret
= btrfs_commit_transaction(trans
, root
);
1058 return ret
> 0 ? EIO
: ret
;
1061 static struct vm_operations_struct btrfs_file_vm_ops
= {
1062 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
1063 .nopage
= filemap_nopage
,
1064 .populate
= filemap_populate
,
1066 .fault
= filemap_fault
,
1068 .page_mkwrite
= btrfs_page_mkwrite
,
1071 static int btrfs_file_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
1073 vma
->vm_ops
= &btrfs_file_vm_ops
;
1074 file_accessed(filp
);
1078 struct file_operations btrfs_file_operations
= {
1079 .llseek
= generic_file_llseek
,
1080 .read
= do_sync_read
,
1081 .aio_read
= generic_file_aio_read
,
1082 .splice_read
= generic_file_splice_read
,
1083 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1084 .sendfile
= generic_file_sendfile
,
1086 .write
= btrfs_file_write
,
1087 .mmap
= btrfs_file_mmap
,
1088 .open
= generic_file_open
,
1089 .release
= btrfs_release_file
,
1090 .fsync
= btrfs_sync_file
,
1091 .unlocked_ioctl
= btrfs_ioctl
,
1092 #ifdef CONFIG_COMPAT
1093 .compat_ioctl
= btrfs_ioctl
,